This is a specialized language model trained on NASA research data, designed to assist with space science questions, research summarization, and technical explanations. The model is based on Phi-3 Mini 3.8B and has been optimized for scientific accuracy and detailed explanations.

Model Details

Base Model: Microsoft Phi-3 Mini 3.8B
Training Data: NASA research papers, mission reports, and scientific publications
Model Size: ~3.8B parameters
License: MIT
Format: GGUF (compatible with Ollama, llama.cpp)

Training Data Sources

NASA research papers from PubMed Central
NASA Taskbook project descriptions
NASA Open Science Data Repository
Space biology and microgravity research
Astronaut health and safety studies
Planetary science publications

Capabilities

Question Answering: Detailed responses about space science topics
Research Summarization: Condensing complex scientific papers
Technical Explanations: Breaking down aerospace concepts
Mission Analysis: Discussing NASA missions and findings
Scientific Accuracy: Trained on peer-reviewed research

Usage

With Ollama

bash

# Pull the model
ollama pull bhavyasri044/ollama-nasa-model

# Run the model
ollama run bhavyasri044/ollama-nasa-model

With llama.cpp

bash

# Download the GGUF file
wget https://huggingface.co/bhavyasri044/ollama-nasa-model/resolve/main/model.gguf

# Run with llama.cpp
./main -m model.gguf -p "What are the effects of microgravity on human bone density?"

Example Queries

"What are the main health risks for astronauts on long-duration missions?"
"Explain the effects of microgravity on plant growth"
"Summarize recent findings about space radiation exposure"
"How does the ISS maintain its orbit?"

Model Performance

The model has been optimized for:

Scientific accuracy in space-related topics
Detailed explanations suitable for researchers and students
Proper citation of NASA research when applicable
Clear communication of complex concepts

Limitations

Knowledge cutoff based on training data (up to 2024)
Primarily focused on NASA and US space research
May not have complete coverage of very recent developments
Should not be used for mission-critical decisions without verification

Training Process

Data Collection: Gathered NASA research papers and publications
Data Processing: Cleaned and chunked scientific texts
RAG Integration: Built retrieval-augmented generation system
Fine-tuning: Applied LoRA fine-tuning on instruction pairs
Optimization: Configured for scientific accuracy and detail

Technical Specifications

Context Length: 4096 tokens
Temperature: 0.7 (balanced creativity/accuracy)
Top-p: 0.9
Quantization: Q4_K_M (recommended for most use cases)

Citation

If you use this model in your research, please cite:

bibtex

@misc{nasa-ollama-model-2024,
  title={NASA Research Assistant: Specialized Language Model for Space Science},
  author={NASA Research Team},
  year={2024},
  publisher={Hugging Face},
  url={https://huggingface.co/bhavyasri044/ollama-nasa-model}
}

Contact

For questions about this model or to report issues, please open an issue on the repository.

Acknowledgments

NASA for providing open access to research data
Microsoft for the Phi-3 base model
The open-source community for tools and libraries

This model is designed for educational and research purposes. Always verify critical information with official NASA sources.